Reorder Recommendations

Shopify Logistics

2022 - 2023


During my time at Shopify I worked alongside the Inventory Management and Optimization team in the logistics division. Proper inventory management is crucial for merchants to entrust Shopify as their fulfillment provider rather than going with a third party, so delivering top notch experiences was critical to the logistics business. This project was born as a response to a common struggle small and medium size businesses encounter everyday when figuring out how much inventory to reorder from their suppliers and send into the Shopify fulfillment network.


Historically, small and medium size merchants have struggled to reorder the correct amount of inventory due to lack of expertise and resources to leverage data for demand planning and replenishment. Shopify merchants were facing frequent out of stocks which would lead to significant amounts of lost sales, as well as overstock which would lead to opportunity cost from capital invested in dead or overstocked inventory as well as high storage fees. Our inventory optimization team along with the data science team took on the challenge to leverage our merchant’s sales and seasonality data to provide them with an exceptional forecast and recommendations and take this time consuming task off of their shoulders so they could focus on growing their businesses.


Sales lost due to out-of-stocks in 2021


Retail value of overstocked items + $711K paid in storage costs


Merchants felt they could use help with forecasting



Project goals

Eliminate the complex and time consuming work Shopify merchants face daily to keep the right amount of inventory in stock by providing them with smart replenishment recommendations so they can focus on what they love.

My role

I led the research and design of the replenishment recommendations experience working alongside the product, engineering and data science team.

Intended outcomes

  • Merchants see reduced out of stocks and increased sales
  • Merchants have improved coverage because there is sufficient inventory in the network
  • Merchants have reduced overhead with recommended SKUs and quantities to share with their suppliers
  • Merchants find value in the recommendations
  • SFN learns about the accuracy of supplier inputs and gains further insight into the critical elements of the reorder recommendation

Target merchants

Intended for general use by all Shopify merchants using the fulfillment service, but we started out by addressing the needs of merchants with the following characteristics:

  • unsophisticated forecasts
  • 1-2 suppliers, and/or low SKU counts
  • D2C service only
  • Shopify storefront only
  • No ERP integration needs



I began immersing myself in the demand planning world so I could speak the same language as the data science team responsible for creating the forecasting model for our merchants. I learned all about reorder points, lead times, safety stock, MOQ’s (minimum order quantities), stockout tolerance, and many more terms I was not fully familiarized with until joining this project.

I spent countless hours asking the data science team how to read forecasting graphs so I could then talk to merchants and learn what they most value in these forecasting models.



I also familiarized myself with the working prototype the data science team had built in google sheets to test with real merchant data the accuracy of the model. 

The data science team had done some light user testing on this prototype with 3 merchants before, so I had some initial merchant feedback to review and use as a starting point.



Next step was to talk to merchants and learn how they were currently doing demand planning and forecasting future sales. Five merchants walked me through their personal spreadsheets and I learned what similarities and differences they had based on their different needs. I also interviewed a few of our internal account management team members to learn how merchants had been requesting their help in the past to view their inventory levels for planning purposes. 


  • Merchants were having to go into multiple different places to search for information
  • We were not providing basic data visibility for them to easily digest and use for their own personal spreadsheets.


We had an ideation workshop with a few more designers where we reviewed the context and problem statement and then had 10 minutes to sketch ideas and later share and explain our sketches. This helped me define a first concept design to bring to the product, engineering and data science team.



Following the initial workshop, I met with product, engineering, and data science teams to assess the feasibility of our ideas. Surprisingly, I discovered a limitation in the model: it couldn't predict future reorder points; instead, it could only alert merchants when items reached their reorder points.



I put together a prototype to test out the “reorder now” concept instead of showing future reorder dates. The prototype included a first time experience where merchants were prompted to import their supplier data through a CSV file, and see a summary card on their inventory page with the number of SKUs being recommended to reorder. These SKUs would be grouped in a new index where it would include a reorder frequency filter as well as relevant data to the recommendation, such as 'Reorder quantity', 'Current weeks of supply' and 'New weeks of supply'.



After one round of user testing with 5 merchants, the team and I were surprised at the results. Merchants hated this proposal, and we learned how important it is for them to fully own the decision of reordering. What they needed most at this moment was visibility into past sales and future forecasts all in one place so they could make sure our recommendation was credible. We had underestimated their trust in our recommendation model, so we pivoted to fully focus on transparency.


“Show me how you got this number”

“Why should I reorder x amount?”


After analyzing merchant feedback from the user testing sessions we learned that merchants own the decision of replenishment and bear the cost. In order to optimize costs for the merchants in the long term, we needed to build trust along that path. So that's how we re-focused our reorder experience - make it easy and make it transparent.

With our data scientist and engineering lead, I led an ideation workshop to define what our merchants most valued in forecast visualization, now with a much more clear picture of what their needs and requirements are. 


I then created the first concepts of the forecast graph to add to the Product detail pages of each SKU with the goal of giving merchants the full picture of our recommendations. I paired with the data-viz Polaris (Shopify’s design system) team to check feasibility and possibly develop new components for our sophisticated and interactive graphs we had in mind.



Along with product and engineering, we had a series of working sessions to redefine scope. We revisited our goals, use cases and merchant journeys and decided to launch in 2 phases:

V1: Focusing on showing merchants which products need replenishment soon and why.

V2: Providing merchants with an accurate recommendation of when and how much to reorder.



Inventory index will now show forecast of remaining weeks of inventory.


Product details page will have a new Forecast card, explaining how we calculate “weeks of supply”.


From the product details page Forecast card, merchants will be able to click through the “View forecast graph” link to have full visibility into their past 5 weeks sales and upcoming 12 weeks forecast, along with their expected stock-out date.



We were able to launch V1 a month before Shopify announced the sale of Shopify Logistics to Flexport. During that time we had good feedback from our merchants around the weeks of the supply column, but did not have enough time to fully measure success in terms of how much V1 had impacted out of stocks and overstock levels. We were also unable to launch V2 which would show the actual reorder quantity and date using supplier data, so results were inconclusive. After the acquisition, this project was deprioritized until further notice.


Although the Flexport acquisition of Shopify Logistics gave this project an ending that was not what I would have expected from a project we dedicated many months to design and build with the team, I am very proud of the process and how we got to the final solution. The more I learned about demand planning from our merchants as well as internal teams, the easier it became to find a solution to a problem that started out very ambiguous. The user testing sessions helped reformulate the product vision scope and because product and engineering participated actively in these sessions, it was easy to get buy-in from the team to reiterate so we could give our merchants the transparency they needed to trust our forecast model.

Case Studies